Fable
The winery tastes the first sip and the guest tastes the first pot
In the south of the city there was a century-old winery, and the winemaker had an ancestral rule: when new wine came out of the cellar, first pour a small cup and place it by the stove. After the time of two incense sticks, smell it for sourness and look for sediment. This was called the pre-stove test. Only if the pre-stove test showed no problem did they dare open the jar and fill the pots.
After opening the branch, the estate owner added a rule: for each brewed jug of wine, first send five jugs to the branch's most familiar regulars' tables, free of charge, and let the guests drink. After drinking, ask three questions: is the taste right, is the aftereffect strong, and does it cause a headache the next day. This is called the table trial. Only after the table trial is fully satisfactory is the new wine's label hung on the wine list the next day.
One year something went wrong. The new wine passed the stove-side check—standing by the stove, it looked clear and had no sour taste. They skipped the table test entirely and hung up the sign, putting it on shelves citywide the next day. The first day's customers all said it was good. By the third day, customers started coming back to return the wine—this batch gave them a headache and dry mouth the next day. What two sticks of incense by the stove could not detect, the customers detected after drinking and sleeping it off.
The estate owner closed the doors for three days and wrote the pre-stove test and the pre-table trial as two separate lines. The pre-stove test checked the stability of the wine itself—whether it would spoil, whether it had impurities, whether the alcohol content was correct. The pre-table trial checked the guest's experience after drinking—mouthfeel, aftereffects, and the body's reaction the next day. The two-incense-stick window and the overnight window were measuring completely different things.
As it grew, the estate owner realized there was another boundary to manage: how many jars should the kitchen-side test cover? Test one jar, and the other ninety-nine might have the same flaw. How many guests should the table-side test cover? Send five jugs to old acquaintances, and their tastes are fixed—they won't catch new problems. The owner raised the kitchen-side test from one jar to a sample of five, and the table-side test from five jugs to thirty, sent to different types of guests. Coverage and sampling strategy also had to be managed separately.
An out-of-town wine merchant came to learn. At the welcome banquet the estate owner drew a diagram: the check at the stove tests whether it runs—whether the wine keeps after being put in the jar. The check at the table tests whether it drinks well—whether guests are satisfied after actually drinking it. It must run before it can be drunk, but running doesn't guarantee it drinks well. The two things can't be reduced to one.
The winery later hung this diagram in the shipping room: new wine goes on the shelf after passing two gates—one at the stove, one at the table. Skip either gate, and you pay for it at that gate.